Skip to main content

Full text of "Clump morphology and evolution in MHD simulations of molecular cloud formation"

See other formats


Mon. Not. R. Astron. Soc. 000, 000-000 (0000) Printed 12 June 2009 (MN I^TgX style file v2.2) 



Clump morphology and evolution in MHD simulations of 
molecular cloud formation 



R. Banerjee 1 , E. Vazquez-Semadeni 2 , P. Hennebelle 3 and R.S. Klessen 1 

1 Zentrum fur Astronomie der Universitdt Heidelberg, Institut fur Theoretische Astrophysik, Albert- Ueberle-Str. 2, 69120 Heidelberg, Germany 

2 Centro de Radioastronomia y Astrofisica ( CRyA ), Universidad Nacional Autonoma de Mexico, Morelia, Michoacdn, Mexico 

3 Laboratoire de radioastronomie millimetrique (UMR 8112 CNRS), 

Ecole Normale Superieure et Observatoire de Paris, 24 rue Lhomond, 75231 Paris Cedex 05, France 



12 June 2009 



ABSTRACT 

We study the properties of clumps formed in three-dimensional weakly magne- 
tized magneto- hydrodynamic simulations of converging flows in the thermally bistable, 
warm neutral medium (WNM). We find that: (1) Similarly to the situation in the 
classical two-phase medium, cold, dense clumps form through dynamically-triggered 
thermal instability in the compressed layer between the convergent flows, and are of- 
ten characterised by a sharp density jump at their boundaries though not always. (2) 
However, the clumps are bounded by phase-transition fronts rather than by contact 
discontinuities, and thus they grow in size and mass mainly by accretion of WNM ma- 
terial through their boundaries. (3) The clump boundaries generally consist of thin 
layers of thermally unstable gas, but these layers are often widened by the turbulence, 
and penetrate deep into the clumps. (4) The clumps are approximately in both ram 
and thermal pressure balance with their surroundings, a condition which causes their 
internal Mach numbers to be comparable to the bulk Mach number of the colliding 
WNM flows. (5) The clumps typically have mean temperatures 20 ^ (T) £ 50 K, cor- 
responding to the wide range of densities they contain (20 £ n £ 5000 cm -3 ) under a 
nearly-isothermal equation of state. (6) The turbulent ram pressure fluctuations of the 
WNM induce density fluctuations that then serve as seeds for local gravitational col- 
lapse within the clumps. (7) The velocity and magnetic fields tend to be aligned with 
each other within the clumps, although both are significantly fluctuating, suggesting 
that the velocity tends to stretch and align the magnetic field with it. (8) The typical 
mean field strength in the clumps is a few times larger than that in the WNM. (9) 
The magnetic field strength in the densest regions within the clumps (n ~ 10 4 cm -3 ) 
has a mean value of B ~ 6 jiG but with a large scatter of nearly two orders of 
magnitude, implying that both sub- and super-critical cores are formed in the simu- 
lation. (10) In the final stages of the evolution the clumps' growth drives them into 
gravitational instability, at which point star formation sets in, and the pressure in the 
clumps' centers increases even further. 

Key words: magneto-hydrodynamics, ISM: clouds, evolution, methods: numerical 



1 INTRODUCTION 

The formation of molecular clouds by converging flows in the 
warm neutral atomic medium (WNM) has been intensively 
studied in recent years through numerical and analytical 



treatments (e.g. 


, pallesteros-Paredes et al. 
Koyama & Inutsuka 2000 


1999||Hennebelle 


& Perault|1999 


Hartmann et al. 


2001||Koyama I 


k Inutsuka 


2002; Audit & 


Hennebelle 2005; 


Heitsch et al.||2005| |Gazo 


et al. 


|2005| Vazquez- Semadeni 


et al. 2006, 2007||Hennebelle et al. 


2007| Heitsch et al.|2008), 



which have shown that molecular regions can form by ther- 



mal instability TI (Field| |1965) in the warm neutral inter- 
stellar medium (ISM), nonlinearly triggered by transonic 
compressions in the WNM. The added ram pressure of such 
compressions causes the affected regions to overshoot from 
cold neutral medium (CNM) to molecular cloud physical 
conditions. However, the nature and evolution of the clumps 
appearing self-consistently within the clouds produced by 
this mechanism remains controversial. On the one hand, the 
clumps have been reported to be in approximate pressure 
balance with their surroundings and to have sharp bound- 



2 R. Banerjee et al 



aries, as in the classical two-phase medium (e.g., Audit & 
Hennebelle 2005 Hennebelle et al.| |2007), while simultane- 
ously, the whole medium exhibits turbulent behavior, with 
wide distributions of the density and pressure. This implies 
the existence of significant amounts of gas in the unstable 
density and temperature ranges, which has been suggested 
to be in transit between the stable warm and cold phases 
(|Vazquez-Semadeni et al.||2000| |Gazol et aL||2001| |Sanchez- | 
Salcedo et al.|2 002 ; Vazquez-Semadeni et al. 2003||de Av illez 



fe Breitschwerdt|2004||Gazol et al.|2005|| Audit Hennebellel 
2005) . 

This coexistence of a two-phase and a turbulent regime, 
to which we refer as "thermally bistable turbulence" , is in- 
triguing, since turbulence implies continuous and irregular 
transport of gas (e.g 



Klessen et al. 2000 Klessen & Lin 



2003), while the two-phase regime is usually thought to im- 



ply static conditions of the clouds, held under confinement 
by the pressure of the warmer, more diffuse WNM, and with 
the two phases being mediated by contact discontinuities at 
pressure equilibrium, which, by definition, imply no fluid 
transport across them. 

In this contribution, we address these issues by means of 
three-dimensional (3D), adaptive mesh refinement (AMR), 
magneto- hydrodynamical (MHD), self-gravitating simula- 
tions, performed with the FLASH code jFryxell et al.|2000| ), 
which allow us to have sufficient resolution to simulate the 
formation of a large dense cloud complex while still resolving 
the interiors of the clumps that naturally form during this 
process. Note that, even though our simulations are mag- 
netic, in this paper we focus on the nature and evolution of 
the clumps, rather than on the effect of the presence of the 
magnetic field, a task that we defer to a future study. 



2 NUMERICAL MODEL 

We model the convergence of WNM flows as two colliding, 
large-scale cylindrical streams, whose evolution we follow 
with the FLASH code under ideal MHD conditions. Our 
setup is similar (though not identical) to the non-magnetic 
SPH simulation of Vazquez-Semadeni et al. ( 2007 ) labeled 



L256Av0.17 (see their Fig. 1). Each stream is 112 pc long 
and has a radius of 32 pc. They are embedded in a (256 pc) 3 
simulation box. Although the numerical box is periodic, the 
cloud occupies a relatively small volume far from the bound- 
aries, and so the cloud can interact freely with its diffuse en- 
vironment, with relatively little effect from the boundaries. 

The cylindrical streams are given an initial, slightly su- 
personic inflow velocity so that they collide at the centre of 
the numerical box. The inflow speed of each stream cor- 
responds to an isothermal Mach number of 1.22, where 
the initial temperature of the atomic gas is 5000 K imply- 
ing an isothermaQ sound speed of 5.7 km s _1 . We also 
add 10% random velocity perturbations to the bulk stream 
speeds. The initially homogeneous density is n = 1 cm 
(p = 2.12 x 10 
1.27). 



24 g cm 3 , using a mean atomic weight of 



1 Note that our inflow speeds are a factor of (5/3) _1 / 2 smaller 
than the one in Vazquez-Semade ni et aT] ([2007) as we use the 
isothermal sound speed, while those authors used the adiabatic 



Here, we report on the results from our weakly mag- 
netised case with a homogeneous magnetic field compo- 
nent of strength B x = 1 /iG, which is parallel to the gas 
streams. These initial conditions translate to a plasma f3 — 
Pthevm/Pmag of 17.34. We choose this relatively weak field 
so that the gas in the streams is magnetically supercritical, 
and thus the cloud formed by compression along the field 
lines can eventually also become supercritical. Note that the 
only way in which our clouds can become supercritical is by 
accreting mass from the inflows alo ng the field lines (|Mes- 



tel|1985||Hennebelle fe Perault|2000l |Hartmann et al.| 2001 
Shu et al.|2007||Vazquez-Semadeni|2007| , since we do not in- 
clude ambipolar diffusion in our simulations. In any case our 
choice of the initial uniform field is not overly low, as recent 
estimates of the mean Galactic field give upper and lower 
limits of 4 ± 1 and 1.4 ± 0.2 fiG, respectively ( |Beck||2001 ). 
On the other hand, total magnetic field strengths which also 
comprise of field fluctuations are typically reported to be of 



the order of 6 fiG (see e.g., Heiles k, Troland 2005). Here, 



we are not including initial field fluctuations. In our simu- 
lation, such fluctuations are dynamically generated in the 
CNM from the initial turbulent velocity and thermal insta- 
bility. 

This setup is different from the numerical setup in our 
companion paper, Hennebelle et al. (2008), where a sig- 



nificantly stronger field was used (5/iG), and the colliding 
streams entered the simulation box along the field direction 
through two opposite boundaries, which had inflow rather 
than periodic conditions. Thus, in that paper, the mass-to- 
flux ratio of the cloud was allowed to increase without limit, 
while in our simulations, this ratio is bounded from above 
by the mass-to-flux ratio of the simulation box. 

These initial conditions result in a ratio of the total 
mass in the flows, Mfl OW s ~ 2.26 x 10 4 Mq, to the ther- 
mal Jeans mass of M flows /Mj « 2 x 10" 3 . The warm 
neutral medium is far from being gravitationally unsta- 
ble. On the other hand, the mass-to-flux ratio of the flows, 
[i — Mflowg/^s = Y,/B x , is such that /x//x C rit ~ 2.9, where 
Merit ~ 0.1 3/ VG is the critical value for a slightly flattened 
structure on the verge of collapse ( |Mouschovias Spitzer 
1976 Spitzer 1978]). Using this critical value, the mass-to- 
flux ratio of the entire simulation box is 3.3/i C rit, because 
the extend of the flows is slightly smaller than the box size. 
We estimate the transition time at which the cloud becomes 
magnetically super-critical as 



t t 



5.4 Myr 

X Ucm- 3 ; V6.9km s" 1 / V 1 



B x 

TJIg 



(i) 



Concerning the cooling, we use the procedure described 
in | Vazquez-Semadeni et ah] ( |2007| , which is based on the 
chemistry and cooling calculations of Koyama & Inutsuka 
( 2000 ) and the analytic fits to them by Koyama & Inutsuka 



(2002|. Thermal conduction is neglected. 

We follow the gravitational collapse with up to 11 AMR 
refinement levels, which correspond to a maximum resolu- 
tion of 8192 grid points^] or a grid spacing of Ax = 0.03 pc 
in each direction. For the dynamical mesh refinement we use 
a Jeans' criterion, where we resolve the local Jeans' length 



2 Note, each refinement level corresponds to a block 8 3 grid points 



Clump structure in cloud formation simulations 3 



with at least 10 grid cells (see Truelove et al.|[l997[ for the 
necessary criterion to prevent artificial fragmentation). It 
should be noted that some authors have advocated that the 
resolution criterion should be to resolve the Field length 



( Field] 1965 ) with at least three grid points ( |Koyama In 



utsuka||2004| |Gressel| |2009). Other authors ( |Hennebelle 
Audit 2007 ) have suggested that the proper scale that needs 
to be resolved is the "sound crossing scale", the product 
of the sound speed and the cooling time. In the warm and 
cold atomic phases, either of these criteria may actually be 
more stringent than the Jeans one but, as also discussed by 
Hennebelle &; Audit (2007), the main effect of insufficient 
resolution in this kind of simulations is to truncate the frag- 
mentation at the smallest resolved scales, in turn causing 
the the low-mass end of the clump mass spectrum to be also 
truncated. However, in this paper we are not concerned with 
resolving small scale fluctuations nor in the detailed shape 
and extension of the core spectrum, but rather with the 
mechanism of clump growth, and the clumps' substructure, 
once they have reached well-resolved sizes. As discussed by 
Vazquez- Semadeni et al. ( 2006 ), the physics of clump growth 



are simple, and do not require a very high resolution. In that 
paper, it was also concluded that, since clumps grow, they 
are necessarily unresolved during the initial stages of their 
formation, but later become well resolved, as they reach suf- 
ficiently large sizes. Thus, we can investigate the substruc- 
ture of clumps that have had enough time to become well 
resolved. Nevertheless, it could be that by also resolving 
small scale fluctuations the conversion of the WNM to the 
CNM might be slightly faster if the surface-to-volume ratio 
of the cold gas increases in this case. On the other hand, 
since we plan to use the present simulations to study the 
star formation process in future contributions, it is impor- 
tant to satisfy the Jeans criterion. In any case, our clumps 
are typically resolved with at least the 10th level of refine- 
ment, corresponding to a resolution of 0.06 pc. 



1995 



Finally, Lagrangian sink particles (see e.g., Bate et al. 
Jappsen et al. 1120051) are created if the local density 



exceeds n > n s i n k = 2 x 10 cm and this location is a 
local minimum of the gravitational potential. These parti- 
cles interact only gravitationally with the gas and with each 
other. They are free to move within the simulation box inde- 
pendent of the underlying mesh (i.e Lagrangian particles). 
The sink particles have an accretion radius of 0.47 pc corre- 
sponding to roughly one Jeans' length at n s i n k- Within this 
accretion radius, gas in excess of n s i n k is deposited into the 
sink if this gas is gravitationally bound. The sink mass in- 
creases acordingly. Note that the initial mass of the sink is 
computed with the dynamically accreted mass; i.e., only the 
mass in excess of n s i n k contributes to the initial sink mass. 



In a forthcoming study we investigate the influence of 
ambipolar diffusion (AD) on molecular cloud formation in 
colliding flows, where we are using the implementation of 
ambipolar diffusion for the FLASH code by Duffin fc Pu-| 
dritz (2008). Ambipolar diffusion might regulate the mag- 



netic field strength in the condensations that are induced 
by thermal instability ( Inoue et al.||2007 Inoue & Inutsuka 
5008). 



3 RESULTS 

3.1 Global features 

As is already well known, the collision of transonic, con- 
verging flows initially produces moderate compressions on 
the linearly stable WNM at the collision front, which are 
sufficiently strong to nonlinearly trigger thermal instabil- 
ity ( |Hennebelle &; Perault|[l999| ), so that the gas rapidly 
cools to temperatures well below 100 K, forming a thin sheet 
that then fragments into filaments and ultimately into small 
clumps. Moreover, the thermal pressure of the dense gas is 
in close balance with the thermal + ram pressure of the 
WNM outside it, and it is at higher densities and pressures 
than the steady-state CNM, overshooting to typical giant 
molec ular cloud physical (density and temperature) condi- 



tions (Vazquez- Semadeni et al. 



2006, n ~ 100 cm" 



a few tens of Kelvins). This process also causes the newly 
formed dense gas to be turbulent, with a transonic velocity 
dispersion with respect to its own sound speed (|Koyama fe 



[ Inuts uka 2002 ; Heitsc h et al.|2005l |Vazquez-Semadeni et al 
2006; Hennebelle et al. 2008). As in the classical two-phase 



model (Field et al. |1969| )7the clumps in the dense gas are 
bounded by sharp density jumps of roughly a factor of 100. 
We refer to this regime of global turbulence subjected to the 
tendency to form dense clumps by TI as "thermally bistable 
turbulence". In what follows, we indistinctly refer to the 
dense clumps as "molecular" , even though we do not actu- 
ally follow the chemistry in our simulations. 

In this regime, the "molecular cloud" is composed of 
a mixture of diffuse and dense gas, with a significant frac- 
tion of gas in the unstable range, which is in transit from 
the diffuse to the dense phase, thus producing a continuous 
though bimodal distribution of densities and temperatures 



flVazquez-Semadeni et al.|2000||Gazol et al.|200Tj|2005|[Au 
dit fe Hennebe lle 2005). In Fig.^TJ we show the structure of 
the dense cloud from our converging flow simulation about 
5 Myr after the first regions collapse and form stars. In par- 
ticular, from the face-on image one can see that the "cloud" 
is not a homogeneous entity, but rather it is composed of 
dense clumps (which should be mostly molecular) embed- 
ded in somewhat less dense filaments (which may be partly 
molecular and partly atomic) (see also Vazquez- Semadeni 
et al.||2007||Hennebelle et al.|2008[ ). 



3.2 Clump properties 

3.2.1 Clump growth mechanism 

As mentioned above, the clumps are mainly bounded by 
sharp density jumps of roughly a factor of 100. However, 
the clumps are connected to large filaments which build up 
in the intersection region of the colliding flows. The main 
difference between the regime in the simulations and the 



classical two-phase model of Field et al. ( 1969| is that, be- 
cause the clumps are formed by turbulent compressions in 
the WNM rather than by linear development of TI, they can 
accrete mass from distances much larger than the scale of 
the fastest- growing mode of TI in the diffuse medium, which 
is typically small. In a turbulent environment, instead, the 
clumps can accrete from the scales associated with the com- 
pressive motion that forms the clumps. This implies that, for 
turbulence-induced clump formation, the duration of clump 



4 R. Banerjee et al 



t = 22.50 Myr 



40 - 



log(N [cnT a ]) t = 22.50 Myr 

22.0 



20 - 



-20 ■ 



40 



-40 



-20 





X IDC] 



20 



40 




-20 20 

y [pc] 



Figure 1. Column density of the inner region of the molecular cloud viewed edge-on (left panel) and face-on (right panel). In the left 
panel, the large scale flows advance inwards from the right and left sides of the box, leaving a lower-density medium there. In both 
panels, the dots mark the projected positions of the sink particles, i.e. regions of local gravitational collapse. The first regions that show 
active star formation appear at about 17 Myr in this simulation. The molecular cloud is largely inhomogeneous, with the "molecular" 
gas (log(7V/cm -2 ) > 20.5) interspersed within the warm atomic gas. Note that the simulation box is 4 times larger (i.e. 256 pc each side) 
than the area shown here. 



t - 15.00 Myr 



t = 20.00 Myr 




-12 -10 



t = 22.50 Myr 




y [pc] 



-12 -10 -8 

y [pc] 



log(N [cm 2 ]) 




-12 -10 -8 

y [pc] 



Figure 2. Column density evolution of a clump (colored yellow-green) in which a self gravitating, cold, dense core (yellow-red) builds up. 
The clump is embedded in a large-scale, filamentary structure (see also Fig.[l| but is separated from the external WNM (light-blue) by a 
sharp boundary (dark green), although it exhibits a complex, fluctuating substructure. The clump grows mainly by accretion of material 
from the WNM. Cold, dense regions become Jeans unstable and start to form stars (indicated by the black dots). See also Fig. [6] 



growth may be much longer than in the case of linear de- 
velopment, and thus one should expect to generally find the 
clumps in a growing stage, rather than in a quasi- static 
state, as in the final state of the linear development of TL 
In turn, this means that there should generally be a net 
mass flux across their boundaries driven by the ram pres- 
sure of the inflow. This dynamic growth is different to the 
situation in the quasi-static two-phase model, in which any 
mass flux through the fronts (evaporation or condensation; 
e.g., Zeld ovich &; Pikel'Ner|^969 ) occurs as a result of the 
tendency to equalize the thermal pressure between the CNM 
and the WNM. We thus refer to the clumps' boundaries as 
phase transition fronts. 

This mechanism has been studied in detail in one di- 
mension by Hennebelle & Perault (1999) and Vazquez- 



Semadeni et al. (2006). The latter authors gave analytical 



solutions for the expansion velocity of the phase transition 
front that separates the warm and cold gas as a function 
of the bulk Mach number of the inflows (see their fig. 3). 
Typically, those speeds are small, because the gas is tightly 
packaged in the clumps, and so the clump size increases 
slowly. In fact, interestingly, the front speed decreases with 
increasing inflow Mach number because the ram pressure of 
the compression then causes the clump density to be suffi- 
ciently large as to overwhelm the larger accretion rate onto 
the cloud, and the net effect is that the front propagates 
at lower speeds. During their growth, the clumps also ocas- 
sionally coalesce with other nearby clumps, a process that 
enhances their growth rate. 

In Fig. [2] we show the column density evolution of a 
typical clump in our simulation, which exhibits the afore- 
mentioned sharp boundaries and growth, although large 



Clump structure in cloud formation simulations 5 



t - 22.50 Myr 




t - 20.05 Myr 



10 




x [pcj 



y [pc] 



Figure 3. Column density of the edge-on view of the clump 
shown in Fig. [2] The clump clearly originates at the intersection of 
the large scale flows and is connected to a large filamentary struc- 
ture. Perpendicualar to the filament the clump grows through the 
propagation of its phase transition front into the WNM. 



Figure 5. Column density of a clump from the hydro simulation. 
The structure of the clump and its growth by accretion of WNM 
material are very similar to the weakly magnetised case shown 
in Fig. [2] 




Figure 4. Three dimensional structure of the clump shown in the 
last panel of Fig. [2] The density iso-surface is shown for number 
density of 600 cm -3 reveals the complex structure of this clump. 



fluctuations are seen within it as well. A few million years 
later, the clump becomes self- gravitating and starts to pro- 
duce local sites of collapse. We also show an edge-on column 
density view of the clump at t = 22.5 Myr in Fig. [3] and a 
3D iso-density image in Fig. [4] which shows the complexity 
of the dense clump. 

It is important to note that the clumps' growth mecha- 
nism is a simple consequence of the thermal bistability of the 
flow, triggered by the transonic compression in the WNM, so 
it is essentially a thermo-hydrodynamic phenomenon, inde- 
pendent of the presence of the magnetic field. Indeed, Fig. [5] 
shows the density and velocity field for a clump in a non- 



magnetic simulation, showing that the same growth mecha- 
nism occurs there as well. The inclusion of the magnetic field 
does not appear to change its basic action, as the gas simply 
tends to flow along field lines, which in turn are re-oriented 



by the inertia of the flow (cf. sec. 3.2.2) 



3.2.2 Clump internal structure 

To examine the internal structure of the dense clumps, in 
Figs. [6] and [7] we show slices through the clump shown in 
the last panel of Fig. [2] at two different x positions, one 
through the site of its maximum density (Fig. [6]), and the 
other closer to its periphery (Fig. [7]). The four panels of each 
figure respectively show maps of the density, temperature, 
pressure, and magnetic field strength of the clump. The res- 
olution in the interior of the clump is 0.03 pc, while its linear 
dimensions at t = 22.5 Myr are seen to be roughly 3x5 pc, 
suggesting that the clump is well resolved. 

At this point, it is convenient to estimate the expected 
pressure-equilibrium value of the density in the clumps, 
when one includes the ram pressure from the colliding in- 
flows. The density and temperature initial conditions of our 
simulation imply a thermal pressure P t h = 5000 K cm -3 . 
From Fig. 2 of Vazquez- Semadeni et al. (2007), it can be 



seen that the (hydrostatic) cold-phase density corresponding 
to this pressure in our simulation is ~ 150 cm -3 . Since the 
inflow speed of the colliding streams is 1.22 times t he so und 
speed in the WNM, the ram pressure is ~ 1.22 2 / ^/5/3 « 1 
times the thermal pressure^] for a total pressure of P to t = 
-Pram « 10 4 K cm . The pressure-equilibrium density 
of the cold gas at this pressure is seen to be, from that figure, 
n ~ 300 cm" 3 . 



3 Recall the Mach number we use is with respect to the isother- 
mal sound speed, but the calculation needs to be performed with 
the adiabatic one (see also Vazquez-Semadeni et al.|2006 >. 



6 



R. Banerjee et al 




Figure 6. Structure of the clump shown in the right panel of Fig. [2] The images show 2D slices through the densest region of the clump. 
Shown are the density (top left), temperature (top right), thermal pressure (bottom left), and magnetic field strength (bottom right; 
note the linear scale). The arrows in the pressure image indicate the velocity field and, in the magnetic field image, the magnetic field 
vectors. The WNM gas streams into the clump predominately along the magnetic flux lines. Note in the top left panel that the clump 
boundaries (dark green) are generally thin, but on occasions become wide and penetrate deep into the core, causing the transition from 
WNM to CNM to "molecular" gas to be smoother. 



It is then noteworthy that, although the clump is clearly 
separated from the WNM by a sharp boundary (dark green 
in the top left panels of Figs. [6] and [7]), its central part 
(Fig. ^ contains densities ranging from n ~ 20 cm -3 to 
5000 cm -3 , and is seen to be strongly turbulent. Specifically, 
the few-tens-of-Kelvins gas (dark green), which is mostly as- 
sociated with the clumps' thin boundaries, is seen to often 
extend over much wider regions, penetrating deep into the 
clump structure. Presumably, this is gas in transit towards 
cold-phase conditions, whose transition has been delayed by 



the turbulence ( Vazquez-Semadeni et al. 2003). Finally, note 



that the clump contains a few warmer, lower-density and 
lower-pressure "holes", which are closer to having WNM 
conditions. In summary, the turbulence within the clumps, 
perhaps aided by self-gravity (see below), causes strong fluc- 
tuations of about one order of magnitude above and below 
the canonical steady-state value of the CNM density. Within 
this clump, the temperature is T ~ 20 — 50 K, whereas the 



surrounding warm gas is still at T ~ 5000 K. The small 
variability of the temperature within the clump is consis- 
tent with the nearly-isothermal behavior of the gas at those 
densities. Indeed, from Fig. 2 of Vazquez-Semad eni et al.| 
( |2007| ), it is seen that the slope j e of the P vs. p curve for 
the dense gas is ~ 0.8, while an isothermal behavior would 
correspond to 7 e = 1. 

The thermal pressure field is similarly seen to have large 
fluctuations within the clump, in fact exceeding those seen 
in the surrounding WNM. The largest values of the ther- 
mal pressure inside the clump are probably caused by the 
beginning of the gravitationally-contracting phase of these 
regions. Such large thermal pressure fluctuations are a re- 
flection of the turbulent character of the ram pressure in the 
clump's environment. 

The velocity dispersion within this clump is ~ 
0.7 kms -1 . Estimating a sound speed of ~ 0.4 kms -1 for the 
gas in the clump, the implied Mach number is ~ 1.75. Given 



Clump structure in cloud formation simulations 7 




Figure 7. Same 2D images than shown in Fig.[6]but cut through slightly off center (x = 2.5 pc, see also Fig. [3]) . Here the clump properties 
are highly distinct. In particular, the density and temperature contrast compared to the WNM is greatly prominent. Due to the thermal 
bistability of the flow, the clump is almost in pressure equilibrium with its surroundings. 



the slightly softer- than- isothermal equation of state implies 
a slightly larger-than-isothermal density jump in this gas, so 
that, under the effect of turbulence alone, one should expect 
a density contrast ~ 5. The additional density enhancement 
may be attributed to incipient gravitational contraction. In- 
deed, the mass of this clump is ~ 28OM0, while the Jeans 
mass for density n — 200 cm -3 is 150M©, indicating that 
the clump is undergoing gravitational contraction, also in- 
dicated by the fact that it has already formed sink parti- 
cles. This is further illustrated in Fig. [8] which shows the 
mass M and the number of Jeans masses Nj of the gas 
above a given density in the region shown in the rightmost 
panel of Fig. [2] and in Fig. [6] It is clearly seen that the 
clump as a whole is gravitationally unstable, containing a 
couple of Jeans masses. The highest density regions do not 
appear Jeans unstable, but this may be a consequence of 
the fact that this clump has already formed sinks, so the 
sink mass has already been removed from the gas phase. 
On the other hand, the lowest- density gas within the clump 
(n ~ 20-30 cm -3 ) is probably due to interference between 
the condensation process and the turbulence, causing some 



accreting gas from the WNM to not be able to immediately 
undergo the phase transition to the cold phase. 

Concerning the magnetic field, from Figs. [6] and [7| we 
see that it tends to be aligned with the velocity field in 
the dense regions, although it is also highly distorted there 
(recall that the initial configuration had the magnetic field 
parallel to the x axis), a phenomenon already observed in 
the 2D simulations of iPassot et al. (1995). The visual 



impression of alignment is confirmed by the histogram of 
v • B/|v||B|, which is shown in Fig. 11 for all of the dense 
(n > 5 x 10 3 cm -3 ) gas in the simulation. The histogram 
clearly exhibits peaks at 1 and —1, indicating alignment. 
The tangling in the clumps indicates that the field has 
been strongly distorted by the turbulent motions in the com- 
pressed regions. In this case, the alignment with the velocity 
field is probably a consequence of the fact that motions non- 
parallel to the field tend to stretch and align it along with 
them (Hennebelle &; Perault 2000). A similar alignment is 
observed in runs with a more realistic mean field strength of 
3/iG, and including an initial fluctuating component of the 



8 R. Banerjee et al 




Figure 8. Mass M and number of Jeans masses Nj of the gas 
above a given density in the region shown in the rightmost panel 
of Fig. [2] and in Fig. [5] (t = 22.5 Myr). The clump is seen to 
be globally gravitationally unstable (i.e., from its largest scales, 
which correspond to the lowest mean densities. 



Figure 10. Probability density function of the magnetic field 
strength in the high density regions (n > 5 x 10 3 cm -3 ), showing 
a variability of one order of magnitude of the field strength above 
and below a most probable value of (B) ~ 6 /iG at t = 22.5 Myr. 



100.0 



10.0 



=1 




Figure 9. Scatter plot and two-dimensional histogram show- 
ing the magnetic field strength and density of all grid points in 
the simulation at t = 22.5 Myr. The most probable value of the 
magnetic field at a given density (indicated by the locus of the 
rightmost apexes of the contours) is seen to scale with density 
roughly as n 1 / 2 (straight line), although a large scatter of almost 
two orders of magnitude is seen around this mean trend at each 
density. On the other hand, the maximum magnetic field strength 
scales only weakly with the gas density (i.e., B m ax oc n 015 for 
the uppermost contour line). 



field, suggesting that this result may actually be expected 
to apply in actual interstellar clouds. 

Fig.[9]shows a scatter plot (dots) and a two-dimensional 
histogram (contours) of the magnetic field strength versus 
the density for each pixel in the simulation. One can see 
that, from n ~ 100 cm -3 to 10 4 cm -3 , the most probable 
value of the magnetic field, indicated by the locus of the 
rightmost apexes of the contours, appears to scale roughly 
as n 1 / 2 , although with great scatter around this value. 

It is worth noting that at densities n ~ 10 4 cm -3 , the 
mean field strength is (B) ~ 10/iG, although field strengths 
ranging from ~ 1 to ~ 100/iG are observed. This is illus- 
trated in Fig. |10[ which shows the field strength distribu- 
tion in the high density gas only (n > 5 x 10 3 cm -3 ). This 




0.000 

-0.5 0.0 0.5 

cos(0 vB ) 

Figure 11. Histogram of cos 6^5 = v • ~B/vB computed over all 
of the high-density (n > 5 x 10 3 cm -3 ) gas in the simulation, 
where v and B are respectively the magnitudes of the vectors v 
and B at t = 22.5 Myr. The velocity and magnetic field vectors 
are clearly seen to show a strong tendency to be either parallel or 
antiparallel. 



suggests that strongly as well as weakly magnetized cores 
should exist within the clumps, implying that some of the 
non-detections of the field through, for example, Zeeman 
measurements ( |Crutcher et al.|1999 Crutcher|2004| ), should 
actually correspond to very low field strengths, rather than 
to alignment effects that mask the field. 



3.2.3 Statistics of mean clump properties 

In order to study the collective properties of all clumps in 
the molecular cloud, we apply a simple algorithm to identify 
them in our simulation data. We define clumps as connected 
regions with density above a certain threshold. For compar- 
ison, we use two threshold values: nthres = 200 cm -3 and 
^thres = 500 cm -3 . We choose these values because they 
bracket the steady-state value of the density of the CNM 



(cf. sec. 3.2.2). We have found that the clump statistics are 



fairly insensitive to the different density thresholds, suggest- 



Clump structure in cloud formation simulations 9 



1.00 



0.10 



0.01 



1000 



100 
35 

32 V 



E ' 



29 

26 
1.0 



^ 0.1 



10 



ZD 



10.0 



1.0 



=tt= 



0.1 
10000 

1000 

100 



« M 



**** 



o<^V^o* 



"co o v <^> 



>°al %<> ® 

o ^> n 



> o 



So oooo^ 



oo 



o 
o o 



o<x>o O^o 

O A 



o*> o : 



ooo^ 



00 o o 



o o 



oo 
o 



o * 

o o o o^ 

oo 



o 
oo 



% ^ o<> 
• o 0> o> o 



o 

O O * 



'o<*> 
o<>o 



0<> <>a 

o o 



> o 
o % 



o°o 
o 

o 



10 



100 



M M,: 



Figure 12. Statistical properties of the molecular clumps iden- 
tified in the cloud at t = 22.5 Myr. Here, we define clumps as 
connected regions with densities above 200 cm -3 . These auanti- 



100.0 r 



10.0 - 



- 10 J i 




20 

time [Myr] 

Figure 13. Evolution of the total cloud mass (defined as gas 
with density n > 100 cm -3 ) and ratio of this mass to the cloud's 
Jeans mass, showing that the entire cloud rapidly grows to contain 
a large number of Jeans masses. 



ing that our results are not significantly biased by our choice 
of threshold. 

In Fig.^Jwe show some of the averaged internal proper- 
ties of the clumps found in the whole cloud at t = 22.5 Myr. 
The masses of these clumps span the range 2 — 400 Mq at 
this time. Clumps more massive than ~ 200 Mq are close to 
being Jeans unstable (see the top panel of this figure, where 
we show the ratio M/Mj as a function of the clump mass 
M) . Some regions within these massive clumps are collapsing 
and will form stars. It is interesting, however, that the self- 
gravitating clumps as a whole tend to have values of M/Mj 
not much larger than unity. At first sight, this might appear 
contradictory with the notion that molecular clouds contain 
many Jeans masses. However, this apparent contradiction is 
resolved by noting that our entire cloud, formed of many 
clumps, does contain a large number of Jeans masses, as il- 
lustrated in Fig. [13] which shows the total mass of the cloud 
(defined as gas with density n > 100 cm -3 ) and the ratio of 
this mass to its Jeans mass, clearly indicating that, by the 
end of the simulation, the entire cloud contains roughly 100 
Jeans masses. This supports the notion that giant molecu- 
lar clouds may consist of molecular clumps immersed in an 
atomic substrate (e.g., Goldsmith & Li 2005; Hennebelle & 
|Inutsuka|2006| . 

The average density and temperature depend only 
weakly on the clump masses. This reflects the fact that the 
filling factor of the higher-density gas within the clumps is 
very low (i.e., most of the mass is at the lower densities), 
and thus the mean density of the clumps is always very 
close to the threshold density for defining them, with the 
most massive clumps having slightly larger mean densities 



( Vazquez- Semadeni et al.|l997|. The near constancy of the 



clumps' densities and temperatures is also indicated by the 
almost linear relation between the clumps' masses and their 
ratio of mass to Jeans mass, as shown in Fig. |12[ 

The velocity dispersion a in the clumps increases with 
clump mass, which, together with the fact that the clumps 
have roughly the same density, implies that a increases with 
size. However, note that the dynamical range of the clump 
sizes is less than one decade and the scatter in the velocity 
data is quite large. This prevents us from fitting a proper 



10 R. Banerjee et al 



<j-L relation here. Estimating a typical sound speed of 
~ 0.4 km s _1 we find that most of the clumps are sub- or 
transonic. Only the most massive clumps, which are already 
in the state of collapse, develop larger supersonic velocities, 
suggesting that such velocities are the result of the gravita- 
tional collapse of the clumps, or regions within them (the 
cores) . 

The fact that the turbulence in the clumps shortly af- 
ter their condensation is transonic, just as is the turbu- 
lence in the surrounding WNM, is remarkable. It appears 
to be consequence that the ram pressure fluctuations in 
the cold gas are excited by the ram pressure of the warm 
gas, i.e., pw^w 00 pcVc- In addition, the cold, dense clumps 
evolve quickly into thermal- pressure equilibrium with their 
warm surroundings (due to the thermal bistability) , so that 
p w Cw ~ pcCc- The two conditions combined imply that the 
thermal Mach numbers in both media are comparable (i.e., 

M w - M c ). 

The typical mean field strength in the clumps exhibits 
an interesting dichotomy: clumps with M £ IOOMq have a 
mean field strength B ~ 2 /iG, independent of the clump's 
mass, while clumps with M ^ IOOMq, of which more than 
half have already formed sinks, have systematically larger 
field strengths of nearly twice that amount. This result is in 
agreement with observational results that the field strength 
is essentially independent of gas density for the WNM and 
CNM, and only begins to increase with density at higher 
densities, where presumably gravitational contraction is at 
work (e.g. |Crutcher et al.|2003| jHeiles fe Troland||2005| ). It 
is nevertheless interesting that the typical field strength for 
the low-mass clumps is roughly twice the mean value for 
the whole simulation, indicating that some amount of mean 
field amplification exists in the clumps with respect to the 
WNM, even if by only a factor of ~ 2. 

Finally, we note that almost all clumps with mass above 
10 Mq show a critical or super-critical mass-to- flux ratio, 
/i, with a mass dependence of ji oc M 0- 25-0-4 (^although this 
result may be an artifact of the low degree of magnetization 
of our simulations (recall the mean field of our magnetized 
simulation is 1/iG). 



3.2.4 Evolution of the cloud 

While the individual clumps inside the molecular cloud grow 
and merge, the cloud continues to accrete mass from the 
WNM (see Fig. 



14 



for the cloud mass evolution). Eventu- 
ally, this leads to the global contraction of the entire cloud. 
In our simulation this happens at t ~ 20 Myr which is about 
15Myr after the first dense clumps have formed. The in- 
creased gravitational potential in the centre of the cloud 
further compresses the gas, therefore converting an increas- 
ing amount of diffuse gas into the dense phase. This relieves 
concerns that the accumulation length required to attain 
molecular cloud-like column densities from one-dimensional 
accumulation of WNM gas is exceedingly long (~ 1 kpc; 




10 20 30 

time [Myr] 



40 



Figure 14. Mass evolution of the dense gas (i.e. n > 100 cm -3 , 
M gas ), sink particles (M s i n k s ,) and their sum (Mtot)- The mass 
accretion rate of the collapsing regions increases from 10 -4 to 
10 -3 MQyr -1 during the cloud evolution. Note that our model 
does not include feedback effects which should ultimately limit 
the star formation efficiency (e.g. see Vazquez-Sem adeni et al.] 
2007 for an estimate of the time at which the clould could be 
disrupted by the formed OB stars.) 



density increase is provided by the three-dimensional grav- 
itational compression of the gas. The importance of lateral 
gravitational contraction was already pointed out in |Hart-| 



mann et al. (2001 ) and was recently also confirmed in three 



dimensional simulations by|Heitsch &; Hartmann| (|2008| (see 



also Dobbs et al. 2008| . The one-dimensional compression 



only provides the cooling and compression necessary for self- 
gravity to become important, which then provides the re- 
maining necessary compression (see also Elme green||2007| . 

The late global contraction finally leads to enhanced 
gas densities and large (~ 10 pc), dense, coherent, and col- 
lapsing regions of mainly "molecular" gas, in which local 
collapse events occur before the global collapse is completed 
(e.g. [Klessen||2001| [Mac Low fe Klessen||2004| ). Eventually, 
a high-density and high-velocity dispersion region forms in 
the overall minimum of gravitational potential, when the 
global collapse finally reaches center. We expect that this 
stage may result in the conditions where massive stars can 
form (e.g.,|Zinnecker &; Yorke 2007 Vazquez-Semadeni et al 
|2008j[2009|. 



The global evolution of our fiducial magnetic simula- 
tion, with a mean field strength of 1/iG, is in general simi- 
lar to the evolution of the non-magnetic fiducial simulation 
in I Vazquez-Semadeni et al.| ( |2007[ run L256Av0.17 there), 
in the sense that the collision of the WNM flows produces 
a CNM cloud that, due to the combined action of com- 
pression and cooling, becomes gravitationally unstable and 
begins to contract and undergo collapse at localized sites. 
Also similar are the evolutions of the dense-gas and stel- 
lar masses (compare Fig. [14 



to fig. 5 of Vazquez-Semadeni 



McKee fc Ostriker|2007| sec. 2.3) , since much of the column |et al.|2007| , and the formation of a ring at the periphery of 



4 We use the projected area of the clump, A yz , and the averaged 
normal field component, (B x ) = V~ x J dV B x , where V is the 
volume of the clump, to calculate the mass-to-flux ratio, i.e. \i = 
M c \ ump /A yz {B x ). 



the cloud ( |Burkert &; Hartmann| 2004 ) . Subtle differences, 
however, do exist beteen the two simulations, due mainly to 
the presence of the magnetic field, albeit weak, and to the 
slightly weaker inflow speed of our simulation compared to 
that of run L256Av0.17. In particular, in the present sim- 
ulation, the global collapse of the ring occurs at a signifi- 



Clump structure in cloud formation simulations 11 



cantly later time than in run L256Av0.17 (t ~ 40 Myr vs. 
t ~ 23 Myr, respectively). This is possibly a consequence of 
the lower inflow speed, which does not cause as strong an 
ejection of material in the radial direction, as well as of the 
presence of the magnetic field, which is perpendicular to this 
ejection direction, both allowing a greater concentration of 
material in the central parts of the cloud, and a lower mass 
of the ring. As a consequence, in our simulation the first 
local collapse events occur in the central parts of the cloud, 
while in run L256Av0.17 they occurred in the peripheral 
ring. 



4 SUMMARY AND CONCLUSIONS 

In this paper we have reported results on the physical prop- 
erties of the dense gas (which we refer to as "molecular") 
structures formed by transonic compressions in the diffuse 
atomic medium, using 3D MHD simulations including self- 
gravity, and radiative heating and cooling laws leading to 
thermal bist ability of the gas. We have defined the clumps 
as connected regions with densities n > 200 cm -3 , which 
selects the clumps formed by a phase transition from the 
diffuse to the dense phases of thermal instability (TI). We 
do not consider in our statistics the substructures within 
these clumps, which would correspond to dense molecular 
cores. 

The ram pressure from the accretion of WNM gas into 
the clumps contributes a net ram pressure, in addition to 
the thermal pressure of the WNM, causing the clumps' den- 
sities to overshoot past the typical conditions of the CNM 
(n ~ 50 cm -3 ), well into the realm of physical conditions 
typical of large molecular clouds (n ^ 200 cm -3 ). Moreover, 
since the ram pressure from the diffuse medium is turbulent 
and fluctuating, it induces transonic turbulence within the 
clumps which, as a consequence of the joint conditions of 
ram and thermal pressure balance, must have an rms Mach 
number comparable to that in the diffuse gas. The transonic 
turbulence in the clumps induces significant density fluctu- 
ations, which then provide the seeds for subsequent local 
gravitational collapse as the clumps approach their Jeans 
mass. 

We found that the transition between such clumps and 
the diffuse medium is generally sharp, with both media be- 
ing at roughly the same thermal pressure, similarly to the 
situation in the classical two-phase medium of |Field et al.| 
(1969). However, the clumps contain large density fluctua- 



implying that they are subject to continuous accretion from 
the WNM driven by its ram pressure ( | Ballesteros-Paredes 



tions within them, of up to one order of magnitude above 
and below the nominal pressure-equilibrium density value, 
caused by the presence of thermally unstable gas still in 
transit towards the cold phase on the one hand, and to local 
gravitational contraction on the other. Thus, the boundaries 
of the clumps, which generally consist of thin layers of ther- 
mally unstable gas, often become extended and penetrate 
deep into the clumps. The clumps are nearly isothermal in- 
side, with temperatures in the range ~ 20-50 K, as con- 
sequence of the density fluctuations within the clumps and 
the nearly isothermal equation of state governing the high- 
density gas. 

Another key difference between the classical model and 
the results of our simulations is that the clumps are formed 
dynamically by the compressions in the surrounding WNM, 



et al.iri999; Ballesteros-Paredes 2006). This in turn causes 



the clumps' mass and size to grow in time. Thus, the clumps' 
boundaries are ram-pressure-driven phase-transition fronts 
and clump growth occurs mainly by accretion through their 
boundaries, rather than by coagulation, as was the case in 
earlier models of the ISM (e.g., |Kwan &; Val des 1983J). In 
turn, this mass flux drives the clumps to eventually become 



gravitationally unstable and collapse (see also Gomez et al. 



2007, for an analogous situation in isothermal flows, with 
clumps being bounded by accretion shocks). 

The magnetic field shows a significant level of alignment 
with the velocity field, but also large fluctuations in magni- 
tude and direction inside the clumps, suggesting that it has 
been significantly distorted by the turbulent motions in the 
dense gas. We also find very similar distortions of the mag- 
netic field structure by turbulent motions in the case with 
larger a initial field strengths of 3 /iG. This suggests that gas 
streams and field lines are likely to be aligned in the cases 
of either a weak or a strong magnetic field. 

The molecular clumps and the cloud as a whole are dy- 
namical and evolve with time, with important consequences 
for their ability to form stars (e.g., Mac Low & Klessen 2004 



Ballesteros-Paredes et al.||2007 ). After some 20 Myr of evo- 



lution, some regions have already undergone local collapse 
and started to form stars, while other clumps do not yet 
show signs of star formation, similarly to the suggestion by 



|Elmegreen (2007) for clouds behind the spiral arms of the 
Galaxy. From Fig. 14, we see that by t ~ 28 Myr, roughly 
15-20% of the total mass in the cloud (dense gas + sinks) 
has been converted to sinks. By this time, according to the 



estimates of Vazquez- Semadeni et al. (2007), based on the 
prescription by Franco et al. (1994), the cloud could be de- 



stroyed by the newly formed massive stars. Since SF began 
in the simulation at t ~ 17 Myr, this implies that the stellar 
age spread in the entire cloud should be ~ 10 Myr. Note, 
however, that our entire cloud, with a physical size of ~ 80 
pc is analogous to a giant cloud complex, rather than to an 
isolated cloud. Local, isolated SF sites of sizes ~ 10 pc, have 
smaller age spreads. 

During the evolution of the cloud, global gravitational 
focusing enlarges connected molecular regions in the centre 
of the cloud (see also |Burkert &; Hartman n 2004; Hart mann| 
|&; Burkert||2007| ). At the stage when the global contraction 
reached the center of the cloud, we expect that the condi- 
tions should be reached where massive stars could form (see, 



e.g., IZinnecker k. Yorke|20071 1 Vazquez- Semadeni et al.|2 008 , 



2009). 



We conclude that the formation and evolution of clumps 
in a thermally bistable medium is a highly complex process 
that retains some of the features from the classical two-phase 
model, such as the frequent presence of sharp density dis- 
continuities, which separate the cold clumps from the warm 
diffuse medium, while at the same time exhibiting a much 
more complex structure, consisting of an intrincate filament 
network connecting the clumps, and made up of mainly ther- 
mally unstable gas. Furthermore, the clumps are internally 
turbulent, and thus have density fluctuations of up to one 
order of magnitude even before they become gravitationally 
unstable. The role of the magnetic field appears negligible at 
the relatively low magnetization levels we have considered 



12 R. Banerjee et al 



in this paper. In a subsequent paper, we will consider more 
strongly magnetized regimes, including ambipolar diffusion. 



ACKNOWLEDGEMENTS 

We thank the anonymous referee for useful comments and 
suggestions on our work which helped to improve this pa- 
per. The FLASH code was developed in part by the DOE- 
supported Alliances Center for Astrophysical Thermonu- 
clear Flashes (ASC) at the University of Chicago. Our simu- 
lations were carried out on the Cluster Platform 4000 (Kan- 
Balam) at DGSCA-UNAM and at HLRB II of the Leibniz 
Rechenzentrum, Garching. RB is funded by the DFG under 
the grant BA 3607/1-1. E.V.-S. acknowledges financial sup- 
port from CONACYT grant U47366-F. R.S.K. thanks for 
subsidies from the German Science Foundation (DFG) under 
Emmy Noether grant KL 1358/1 and grants KL 1358/4 and 
KL 1358/5. This work was supported in part by a FRON- 
TIER grant of Heidelberg University sponsored by the Ger- 
man Excellence Initiative, as well as by the Federal Ministry 
of Education and Research via grant 17EcCZXd. 



REFERENCES 

Audit E., Hennebelle P., 2005, A&A, 433, 1 
Ballesteros-Paredes J., 2006, MNRAS, 372, 443 
Ballesteros-Paredes J., Hartmann L., Vazquez- Semadeni 

E., 1999, ApJ, 527, 285 
Ballesteros-Paredes J., Klessen R. S., Mac Low M.-M., 

Vazquez- Semadeni E., 2007, in Reipurth B., Jewitt D., 

Keil K., eds, Protostars and Planets V Molecular Cloud 

Turbulence and Star Formation, pp 63-80 
Ballesteros-Paredes J., Vazquez-Semadeni E., Scalo J., 

1999, ApJ, 515, 286 
Bate M. R., Bonnell I. A., Price N. M., 1995, MNRAS, 277, 

362 

Beck R., 2001, Space Science Reviews, 99, 243 

Burkert A., Hartmann L., 2004, ApJ, 616, 288 

Crutcher R., Heiles O, Troland T., 2003, in Falgarone E., 
Passot T., eds, Turbulence and Magnetic Fields in As- 
trophysics Vol. 614 of Lecture Notes in Physics, Berlin 
Springer Verlag, Observations of Interstellar Magnetic 
Fields, pp 155-181 

Crutcher R. M., 2004, Ap&SS, 292, 225 

Crutcher R. M., Troland T. H., Lazareff B., Paubert G., 
Kazes I., 1999, ApJ, 514, L121 

de Avillez M. A., Breitschwerdt D., 2004, A&A, 425, 899 

Dobbs C. L, Glover S. C. O, Clark P. C, Klessen R. S., 
2008, MNRAS, 389, 1097 

Duffin D. F., Pudritz R. E., 2008, MNRAS in press 

Elmegreen B. G., 2007, ApJ, 668, 1064 

Field G. B., 1965, ApJ, 142, 531 

Field G. B., Goldsmith D. W., Habing H. J., 1969, ApJ, 
155, L149+ 

Franco J., Shore S. N., Tenorio-Tagle G., 1994, ApJ, 436, 
795 

Fryxell B., Olson K., Ricker P., Timmes F. X., Zingale M., 
Lamb D. Q., MacNeice P., Rosner R., Truran J. W., Tufo 
H., 2000, ApJS, 131, 273 



Gazol A., Vazquez-Semadeni E., Kim J., 2005, ApJ, 630, 
911 

Gazol A., Vazquez-Semadeni E., Sanchez-Salcedo F. J., 

Scalo J., 2001, ApJ, 557, L121 
Goldsmith P. F., Li D., 2005, ApJ, 622, 938 
Gomez G. O, Vazquez-Semadeni E., Shadmehri M., 

Ballesteros-Paredes J., 2007, ApJ, 669, 1042 
Gressel O., 2009, ArXiv e-prints 

Hartmann L., Ballesteros-Paredes J., Bergin E. A., 2001, 

ApJ, 562, 852 
Hartmann L., Burkert A., 2007, ApJ, 654, 988 
Heiles O, Troland T. H., 2005, ApJ, 624, 773 
Heitsch F., Burkert A., Hartmann L. W., Slyz A. D., De- 

vriendt J. E. G., 2005, ApJ, 633, L113 
Heitsch F., Hartmann L., 2008, ApJ, 689, 290 
Heitsch F., Hartmann L. W., Slyz A. D., Devriendt J. E. G., 

Burkert A., 2008, ApJ, 674, 316 
Hennebelle P., Audit E., 2007, A&A, 465, 431 
Hennebelle P., Audit E., Miville-Deschenes M.-A., 2007, 

A&A, 465, 445 
Hennebelle P., Banerjee R., Vazquez-Semadeni E., Klessen 

R. S., Audit E., 2008, A&A, 486, L43 
Hennebelle P., Inutsuka S.-L, 2006, ApJ, 647, 404 
Hennebelle P., Perault M., 1999, A&A, 351, 309 
Hennebelle P., Perault M., 2000, A&A, 359, 1124 
Inoue T., Inutsuka S.-L, 2008, ApJ, 687, 303 
Inoue T., Inutsuka S.-L, Koyama H., 2007, ApJ, 658, L99 
Jappsen A.-K., Klessen R. S., Larson R. B., Li Y., Mac 

Low M.-M., 2005, A&A, 435, 611 
Klessen R. S., 2001, ApJ, 556, 837 

Klessen R. S., Heitsch F., Mac Low M.-M., 2000, ApJ, 535, 
887 

Klessen R. S., Lin D. N., 2003, Phys. Rev. E, 67, 046311 

Koyama H., Inutsuka S.-L, 2000, ApJ, 532, 980 

Koyama H., Inutsuka S.-L, 2002, ApJ, 564, L97 

Koyama H., Inutsuka S.-L, 2004, ApJ, 602, L25 

Kwan J., Valdes F., 1983, ApJ, 271, 604 

Mac Low M.-M., Klessen R. S., 2004, Reviews of Modern 

Physics, 76, 125 
McKee C. F., Ostriker E. O, 2007, ARA&A, 45, 565 
Mestel L., 1985, in Black D. O, Matthews M. S., eds, Pro- 
tostars and Planets II Magnetic fields, pp 320-339 
Mouschovias T. O, Spitzer Jr. L., 1976, ApJ, 210, 326 
Passot T., Vazquez-Semadeni E., Pouquet A., 1995, ApJ, 
455, 536 

Sanchez-Salcedo F. J., Vazquez-Semadeni E., Gazol A., 

2002, ApJ, 577, 768 
Shu F. H., Allen R. J., Lizano S., Galli D., 2007, ApJ, 662, 

L75 

Spitzer L., 1978, Physical processes in the interstellar 
medium. New York Wiley-Interscience, 1978. 333 p. 

Truelove J. K., Klein R. I., McKee OF., Holliman J. H., 
Howell L. H., Greenough J. A., 1997, ApJ, 489, L179+ 

Vazquez-Semadeni E., 2007, in Elmegreen B. G., Palo us J., 
eds, IAU Symposium Vol. 237 of IAU Symposium, Molec- 
ular cloud turbulence and the star formation efficiency: 
enlarging the scope, pp 292-299 

Vazquez-Semadeni E., Ballesteros-Paredes J., Klessen 
R. S., 2003, ApJ, 585, L131 

Vazquez-Semadeni E., Ballesteros-Paredes J., Klessen 
R. S., Jappsen A. K., 2008, in Beuther H., Linz H., Hen- 
ning T., eds, Astronomical Society of the Pacific Confer- 



Clump structure in cloud formation simulations 



ence Series Vol. 387 of Astronomical Society of the Pacific 
Conference Series, Massive Star- Forming Regions: Turbu- 
lent Support or Global Collapse?, pp 240 — h 

Vazquez- Semadeni E., Ballesteros-Paredes J., Rodriguez 
L. F., 1997, ApJ, 474, 292 

Vazquez- Semadeni E., Gazol A., Passot T., et al. 2003, 
in Falgarone E., Passot T., eds, Turbulence and Mag- 
netic Fields in Astrophysics Vol. 614 of Lecture Notes in 
Physics, Berlin Springer Verlag, Thermal Instability and 
Magnetic Pressure in the Turbulent Interstellar Medium, 
pp 213-251 

Vazquez- Semadeni E., Gazol A., Scalo J., 2000, ApJ, 540, 
271 

Vazquez- Semadeni E., Gomez G. C, Jappsen A. K., 
Ballesteros-Paredes J., Gonzalez R. F., Klessen R. S., 
2007, ApJ, 657, 870 

Vazquez- Semadeni E., Gomez G. C, Jappsen A. K., 
Ballesteros-Paredes J., Klessen R. S., 2009, arXiv:astro- 

P h 

Vazquez- Semadeni E., Ryu D., Passot T., Gonzalez R. F., 

Gazol A., 2006, ApJ, 643, 245 
Zeldovich Y. B., Pikel'Ner S. B., 1969, Soviet Journal of 

Experimental and Theoretical Physics, 29, 170 
Zinnecker H., Yorke H. W., 2007, ARA&A, 45, 481